Key Points:
Being burned is probably the best way to learn things, and, having launched a lot of sites, I've also been burned a lot. So hopefully this little blog post can avoid some problems for others. Obviously if you have a small, new site, then your approach to launching a site can be very casual. But, especially if you are launching a prominent site, "flipping the switch" on a part of your site that's moved to a new platform, and / or need to launch a site precisely down to the minute, then you need to be a bit more coordinated.
If I were to give just one piece of advice: do as little as possible during the actual launch.
This sounds obvious, but I'm amazed at how much last minute tweaking teams are often assuming for their launch. Let's say you're shooting for a noon GMT site launch and the launch will be on a brand new platform that you've migrated to. The approach perhaps taken the most often? Lump everything into the final hours before the site is launched, and then end up needing to jettison functionality and / or quality at the very last minute.
Another approach would be to take steps such that the only thing you have to do is at noon GMT to change the DNS entry, so you're literally flipping one switch at the end. Again, this sounds completely obvious but is usually overlooked. What are some ways to pull this off?
- Concentrate heavily on the tasks that can occur days before (for example: freeze development, run the search index, migrate the latest comments and content, change the DNS TTL, do final end-to-end testing).
- Also you can focus on what tasks can be started hours before (final search indexing, migration, etc), perhaps by freezing new comments. Then you can do a final test just before launch, with all the content you're going to be starting up with.
What's a good way to see if you are in fact launching with a small to-do list? Conduct a thorough test of the site days before, and rigorously push on why certain features / content are not yet available for testing. This usually means something else needs to be done at the very end, and every last-minute task that needs to be completed increases your risk.
